The Qlik team highly advises ever scrambling (masking, transforming) those data objects which are part of table keys. The reason is because this would potentially destroy any data integrity.
In the specific example of PERNRs, this value is used as a key on literally hundreds of various HR-related tables like those within the range of PA0000-PA9999. If this value were to be masked the value might be 12345678 in one table and 98765432 in another, and 24683579 in yet another. This result would make the data unusable because the dependencies/relationships between the tables would be broken.
The recommendation would be to focus the scrambling instead on the data's attributes; some examples specific to HR: name, address, email, telephone, birthdate, SSN/SIN, salary, bank account info, and family members. Non-HR examples: customer tax IDs, vendor bank account info, vendor tax IDs, vendor VAT number, Business Partner names, and addresses.
Reference the Gold Client Data Transform User Guide for additional relevant content.